This module implements the ability to run a Node.js based JavaScript program as a native Windows or Linux service.
This module is installed using node package manager (npm):
# This module contains C++ source code which will be compiled
# during installation on Windows platforms using node-gyp. A
# suitable build chain must be configured on Windows platforms
# before installation.
npm install os-service
It is loaded using the require()
function:
var service = require ("os-service");
A program can then be added, removed and run as a service:
service.add ("my-service");
service.remove ("my-service");
var logStream = fs.createWriteStream ("my-service.log");
service.run (logStream, function () {
console.log ("stop request received");
service.stop ();
});
Two approaches can be taken when adding and removing services.
In the first approach a program can be responsible for adding, removing and
starting itself as a service. This is typically achieved by supporting
program arguments such as --add
, --remove
, and --run
, and executing the
appropriate action.
The following example adds the calling program as a service when called
with a --add
parameter, and removes the created service when called with a
--remove
parameter:
if (process.argv[2] == "--add") {
service.add ("my-service", {programArgs: ["--run"]}, function(error){
if (error)
console.trace(error);
});
} else if (process.argv[2] == "--remove") {
service.remove ("my-service", function(error){
if (error)
console.trace(error);
});
} else if (process.argv[2] == "--run") {
var logStream = fs.createWriteStream (process.argv[1] + ".log");
service.run (logStream, function () {
service.stop (0);
});
// Run service program code...
} else {
// Show usage...
}
Note the --run
argument passed in the options
parameter to the
service.add()
function. When the service is started using the Windows
Service Control Manager, or the Linux service management facilities, the first
argument to the program will be --run
. The above program checks for this and
if specified runs as a service using the service.run()
function.
Also note that neither the node binary or the programs fully qualified path
are specified. These parameters are automatically calculated it not
specified. Refer to the service.add()
function description for details
about how this works.
In the second approach a dedicated service management program can be
responsible for adding and removing many services in batch. The program
adding and removing services is not a service itself, and would never call
the service.run()
function.
The following example adds or removes number of services:
if (program.argv[2] == "--add") {
service.add ("service1", {programPath: "c:\example\svc1.js",
function(error) {
if (error) {
console.trace(error);
} else {
service.add ("service2", {programPath: "c:\example\svc2.js",
function(error) {
if (error) {
console.trace(error);
}
});
}
});
} else {
service.remove ("service2", function(error) {
if (error) {
console.trace(error);
} else {
service.remove ("service1", function(error) {
if (error) {
console.trace(error);
}
});
}
});
}
Note that unlike the previous example the --run
argument is not passed in
the options
parameter to the service.add()
function. Since each service
program does not add or remove itself as a service it only needs to run, and
as such does not need to be told to so.
Also note that the programPath
argument is passed in the options parameter
to the service.add()
function, to specify the fully qualified path to each
service program - which would otherwise default to the service management
program adding the services.
Each of the service programs can simply start themselves as services using the following code:
var logStream = fs.createWriteStream (process.argv[1] + ".log");
service.run (logStream, function () {
service.stop (0);
});
// Run service program code...
When a service program starts it can always call the service.run()
function
regardless of whether it is started at the console, by the Windows Service
Control Manager, or the Linux service management facilities.
On Windows, when the service.run()
function is called this module will
attempt to connect to the Windows Service Control Manager so that control
requests can be received - so that the service can be stopped. When starting a
program at the console an attempt to connect to the Windows Service Control
Manager will fail. In this case the service.run()
function will assume the
program is running at the console and silently ignore this error.
On Linux, services started at the console will run in the foreground, this
allows command sequences such as CTRL+C
to be used, e.g. during development.
When Linux services are started using the Linux service management facilities,
i.e. service my-service start
, they can be stopped using the signals SIGINT
and SIGTERM
, or again using the Linux service management facilities, i.e.
service my-service stop
.
This behaviour results in a program which can be run either at the console, the Windows Service Control Manager, or the Linux service management facilities with no change.
Upon starting the current working directory of a service program will be
platform specific the , e.g. the "%windir%\system32"
directory on Windows.
Service programs need to consider this when working with relative directory and
file paths.
This path will most likely be different when running the same program at the
console, so a service program may wish to change the current working
directory to a more suitable location using the process.chdir()
function to
avoid different behaviour between the two methods of starting a program.
This module attempts to behave in exactly the same way on Windows and Linux platforms - at least the API is exactly the same for both platforms both from a service management and service running perspective.
On Windows platforms the Windows Service Control Manager WIN32 API is used to
manage services. On Linux platforms a systemd
unit is created if it is
available, otherwise the chkconfig
command is used. If chkconfig
is not
available the update-rc.d
command is tried instead.
The add()
function adds a service.
The name
parameter specifies the name of the created service. The optional
options
parameter is an object, and can contain the following items:
displayName
- The services display name, defaults to thename
parameter- this parameter will be used on Windows platforms only
nodePath
- The fully qualified path to the node binary used to run the service (i.e.c:\Program Files\nodejs\node.exe
, defaults to the value ofprocess.execPath
nodeArgs
- An array of strings specifying parameters to pass tonodePath
, defaults to[]
programPath
- The program to run usingnodePath
, defaults to the value ofprocess.argv[1]
programArgs
- An array of strings specifying parameters to pass toprogramPath
, defaults to[]
runLevels
- An array of numbers specifying Linux run-levels at which the service should be started for Linux platforms, defaults to[2, 3, 4, 5]
, this is only used whenchkconfig
orupdate-rc.d
is used to install a serviceusername
- For Windows platforms a username and password can be specified, the service will be run using these credentials when started, see theCreateService()
functions win32 API documentation for details on the format of the username, on all other platforms this parameter is ignoredpassword
- See theusername
parametersystemdWantedBy
- For when systemd will be used a target can be specified for theWantedBy
attribute under the[Install]
section in the generated systemd unit file, defaults tomulti-user.target
dependencies
- AN array of strings specifying other services this service depends on, this is optional
The service will be set to automatically start at boot time, but not started.
The service can be started using the net start "my-service"
command on
Windows and service my-service start
on Linux.
The cb
callback function is called once the service has been added. The
following arguments will be passed to the callback function:
error
- Instance of theError
class, ornull
if no error occurred
The following example installs a service named my-service
, it explicitly
specifies the services display name, and specifies a number of parameters to
the program:
var options = {
displayName: "MyService",
programArgs: ["--server-port", 8888],
username: ".\Stephen Vickers",
password: "MyPassword :)"
};
service.add ("my-service", options, function(error) {
if (error)
console.trace(error);
});
The remove()
function removes a service.
The name
parameter specifies the name of the service to remove. This will
be the same name
parameter specified when adding the service.
The service must be in a stopped state for it to be removed. The
net stop "my-service"
command can be used to stop the service on Windows and
the service my-service stop
on Linux before it is to be removed.
The cb
callback function is called once the service has been removed. The
following arguments will be passed to the callback function:
error
- Instance of theError
class, ornull
if no error occurred
The following example removes the service named my-service
:
service.remove ("my-service", function(error) {
if (error)
console.trace(error);
});
The run()
function will attempt to run the program as a service.
The programs process.stdout
stream will be replaced with the
stdoutLogStream
parameter, and the programs process.stderr
stream
replaced with the stdoutLogStream
parameter (this allows the redirection of
all console.log()
type calls to a service specific log file). If the
stderrLogStream
parameter is not specified the programs process.stderr
stream will be replaced with the stdoutLogStream
parameter. The callback
function will be called when the service receives a stop request, e.g. because
the Windows Service Controller was used to send a stop request to the service,
or a SIGTERM
signal was received.
The program should perform cleanup tasks and then call the service.stop()
function.
The following example starts a program as a service, it uses the same log stream for standard output and standard error:
var logStream = fs.createWriteStream ("my-service.log");
service.run (logStream, function () {
console.log ("stop request received");
service.stop ();
});
The stop()
function will cause the service to stop, and the calling program
to exit.
Once the service has been stopped this function will terminate the program by
calling the process.exit()
function, passing to it the rcode
parameter
which defaults to 0
. Before calling this function ensure the program has
finished performing cleanup tasks.
BE AWARE, THIS FUNCTION WILL NOT RETURN.
The following example stops the calling program specifying a return code of
0
, the function will not return:
var logStream = fs.createWriteStream ("my-service.log");
service.run (logStream, function () {
console.log ("stop request received");
service.stop (0);
});
Example programs are included under the modules example
directory.
None, yet!
Bug reports should be sent to [email protected].
- Initial release
- Support Linux platforms which don't have the start-stop-daemon program
- Linux start/stop link under
/etc/rcN.d
directories are not removed
- Host repository on GitHub
- Migrate C++ addon code to use the Native Abstractions for Node framework
- Add missing shebang line '#!' to start/stop script template
- Not possible to specify run levels for Linux start/stop script (added new
runLevels
item to theoptions
parameter to theadd()
function
- Remove extraneous semicolon from the README.md file
- On Windows platforms allow users to specify a username/password with which
a service should be run, the
username
andpassword
options parameters were added to theadd()
function
- Require nan 2.3.x to support node version 6
- Support Linux systemd
- The systemd install doesn't work because of typo in directory name in index.js
- Service not automatically started on boot when under the systemd service
(added
WantedBy
attribute to generated systemd unit) - Umask not set in system 5 init script
- Address warnings for 'v8::Value::ToUint32 was declared deprecated'
- Override the stdout/stderr handles instead of using the deprecated
__defineGetter__()
function - Specify dependancies when adding a service
Suggestions and requirements should be sent to [email protected].
Copyright (c) 2014 Stephen Vickers
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
Stephen Vickers [email protected]